replicant-frameworks_native/include/ui/InputReader.h

954 lines
30 KiB
C
Raw Normal View History

Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
/*
* Copyright (C) 2010 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef _UI_INPUT_READER_H
#define _UI_INPUT_READER_H
#include <ui/EventHub.h>
#include <ui/Input.h>
#include <ui/InputDispatcher.h>
#include <utils/KeyedVector.h>
#include <utils/threads.h>
#include <utils/Timers.h>
#include <utils/RefBase.h>
#include <utils/String8.h>
#include <utils/BitSet.h>
#include <stddef.h>
#include <unistd.h>
namespace android {
class InputDevice;
class InputMapper;
/* Describes a virtual key. */
struct VirtualKeyDefinition {
int32_t scanCode;
// configured position data, specified in display coords
int32_t centerX;
int32_t centerY;
int32_t width;
int32_t height;
};
/* Specifies input device calibration settings. */
class InputDeviceCalibration {
public:
InputDeviceCalibration();
void clear();
void addProperty(const String8& key, const String8& value);
bool tryGetProperty(const String8& key, String8& outValue) const;
bool tryGetProperty(const String8& key, int32_t& outValue) const;
bool tryGetProperty(const String8& key, float& outValue) const;
private:
KeyedVector<String8, String8> mProperties;
};
/*
* Input reader policy interface.
*
* The input reader policy is used by the input reader to interact with the Window Manager
* and other system components.
*
* The actual implementation is partially supported by callbacks into the DVM
* via JNI. This interface is also mocked in the unit tests.
*/
class InputReaderPolicyInterface : public virtual RefBase {
protected:
InputReaderPolicyInterface() { }
virtual ~InputReaderPolicyInterface() { }
public:
/* Display orientations. */
enum {
ROTATION_0 = 0,
ROTATION_90 = 1,
ROTATION_180 = 2,
ROTATION_270 = 3
};
/* Actions returned by interceptXXX methods. */
enum {
// The input dispatcher should do nothing and discard the input unless other
// flags are set.
ACTION_NONE = 0,
// The input dispatcher should dispatch the input to the application.
ACTION_DISPATCH = 0x00000001,
};
/* Gets information about the display with the specified id.
* Returns true if the display info is available, false otherwise.
*/
virtual bool getDisplayInfo(int32_t displayId,
int32_t* width, int32_t* height, int32_t* orientation) = 0;
/* Provides feedback for a virtual key down.
*/
virtual void virtualKeyDownFeedback() = 0;
/* Intercepts a key event.
* The policy can use this method as an opportunity to perform power management functions
* and early event preprocessing such as updating policy flags.
*
* Returns a policy action constant such as ACTION_DISPATCH.
*/
virtual int32_t interceptKey(nsecs_t when, int32_t deviceId,
bool down, int32_t keyCode, int32_t scanCode, uint32_t& policyFlags) = 0;
/* Intercepts a switch event.
* The policy can use this method as an opportunity to perform power management functions
* and early event preprocessing such as updating policy flags.
*
* Switches are not dispatched to applications so this method should
* usually return ACTION_NONE.
*/
virtual int32_t interceptSwitch(nsecs_t when, int32_t switchCode, int32_t switchValue,
uint32_t& policyFlags) = 0;
/* Intercepts a generic touch, trackball or other event.
* The policy can use this method as an opportunity to perform power management functions
* and early event preprocessing such as updating policy flags.
*
* Returns a policy action constant such as ACTION_DISPATCH.
*/
virtual int32_t interceptGeneric(nsecs_t when, uint32_t& policyFlags) = 0;
/* Determines whether to turn on some hacks we have to improve the touch interaction with a
* certain device whose screen currently is not all that good.
*/
virtual bool filterTouchEvents() = 0;
/* Determines whether to turn on some hacks to improve touch interaction with another device
* where touch coordinate data can get corrupted.
*/
virtual bool filterJumpyTouchEvents() = 0;
/* Gets the configured virtual key definitions for an input device. */
virtual void getVirtualKeyDefinitions(const String8& deviceName,
Vector<VirtualKeyDefinition>& outVirtualKeyDefinitions) = 0;
/* Gets the calibration for an input device. */
virtual void getInputDeviceCalibration(const String8& deviceName,
InputDeviceCalibration& outCalibration) = 0;
/* Gets the excluded device names for the platform. */
virtual void getExcludedDeviceNames(Vector<String8>& outExcludedDeviceNames) = 0;
};
/* Processes raw input events and sends cooked event data to an input dispatcher. */
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
class InputReaderInterface : public virtual RefBase {
protected:
InputReaderInterface() { }
virtual ~InputReaderInterface() { }
public:
/* Dumps the state of the input reader.
*
* This method may be called on any thread (usually by the input manager). */
virtual void dump(String8& dump) = 0;
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
/* Runs a single iteration of the processing loop.
* Nominally reads and processes one incoming message from the EventHub.
*
* This method should be called on the input reader thread.
*/
virtual void loopOnce() = 0;
/* Gets the current input device configuration.
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
*
* This method may be called on any thread (usually by the input manager).
*/
virtual void getInputConfiguration(InputConfiguration* outConfiguration) = 0;
/* Gets information about the specified input device.
* Returns OK if the device information was obtained or NAME_NOT_FOUND if there
* was no such device.
*
* This method may be called on any thread (usually by the input manager).
*/
virtual status_t getInputDeviceInfo(int32_t deviceId, InputDeviceInfo* outDeviceInfo) = 0;
/* Gets the list of all registered device ids. */
virtual void getInputDeviceIds(Vector<int32_t>& outDeviceIds) = 0;
/* Query current input state. */
virtual int32_t getScanCodeState(int32_t deviceId, uint32_t sourceMask,
int32_t scanCode) = 0;
virtual int32_t getKeyCodeState(int32_t deviceId, uint32_t sourceMask,
int32_t keyCode) = 0;
virtual int32_t getSwitchState(int32_t deviceId, uint32_t sourceMask,
int32_t sw) = 0;
/* Determine whether physical keys exist for the given framework-domain key codes. */
virtual bool hasKeys(int32_t deviceId, uint32_t sourceMask,
size_t numCodes, const int32_t* keyCodes, uint8_t* outFlags) = 0;
};
/* Internal interface used by individual input devices to access global input device state
* and parameters maintained by the input reader.
*/
class InputReaderContext {
protected:
InputReaderContext() { }
virtual ~InputReaderContext() { }
public:
virtual void updateGlobalMetaState() = 0;
virtual int32_t getGlobalMetaState() = 0;
virtual InputReaderPolicyInterface* getPolicy() = 0;
virtual InputDispatcherInterface* getDispatcher() = 0;
virtual EventHubInterface* getEventHub() = 0;
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
};
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
/* The input reader reads raw event data from the event hub and processes it into input events
* that it sends to the input dispatcher. Some functions of the input reader, such as early
* event filtering in low power states, are controlled by a separate policy object.
*
* IMPORTANT INVARIANT:
* Because the policy and dispatcher can potentially block or cause re-entrance into
* the input reader, the input reader never calls into other components while holding
* an exclusive internal lock whenever re-entrance can happen.
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
*/
class InputReader : public InputReaderInterface, private InputReaderContext {
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
public:
InputReader(const sp<EventHubInterface>& eventHub,
const sp<InputReaderPolicyInterface>& policy,
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
const sp<InputDispatcherInterface>& dispatcher);
virtual ~InputReader();
virtual void dump(String8& dump);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
virtual void loopOnce();
virtual void getInputConfiguration(InputConfiguration* outConfiguration);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
virtual status_t getInputDeviceInfo(int32_t deviceId, InputDeviceInfo* outDeviceInfo);
virtual void getInputDeviceIds(Vector<int32_t>& outDeviceIds);
virtual int32_t getScanCodeState(int32_t deviceId, uint32_t sourceMask,
int32_t scanCode);
virtual int32_t getKeyCodeState(int32_t deviceId, uint32_t sourceMask,
int32_t keyCode);
virtual int32_t getSwitchState(int32_t deviceId, uint32_t sourceMask,
int32_t sw);
virtual bool hasKeys(int32_t deviceId, uint32_t sourceMask,
size_t numCodes, const int32_t* keyCodes, uint8_t* outFlags);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
private:
sp<EventHubInterface> mEventHub;
sp<InputReaderPolicyInterface> mPolicy;
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
sp<InputDispatcherInterface> mDispatcher;
virtual InputReaderPolicyInterface* getPolicy() { return mPolicy.get(); }
virtual InputDispatcherInterface* getDispatcher() { return mDispatcher.get(); }
virtual EventHubInterface* getEventHub() { return mEventHub.get(); }
// This reader/writer lock guards the list of input devices.
// The writer lock must be held whenever the list of input devices is modified
// and then promptly released.
// The reader lock must be held whenever the list of input devices is traversed or an
// input device in the list is accessed.
// This lock only protects the registry and prevents inadvertent deletion of device objects
// that are in use. Individual devices are responsible for guarding their own internal state
// as needed for concurrent operation.
RWLock mDeviceRegistryLock;
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
KeyedVector<int32_t, InputDevice*> mDevices;
// low-level input event decoding and device management
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
void process(const RawEvent* rawEvent);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
void addDevice(nsecs_t when, int32_t deviceId);
void removeDevice(nsecs_t when, int32_t deviceId);
InputDevice* createDevice(int32_t deviceId, const String8& name, uint32_t classes);
void configureExcludedDevices();
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
void consumeEvent(const RawEvent* rawEvent);
void handleConfigurationChanged(nsecs_t when);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
// state management for all devices
Mutex mStateLock;
int32_t mGlobalMetaState;
virtual void updateGlobalMetaState();
virtual int32_t getGlobalMetaState();
InputConfiguration mInputConfiguration;
void updateInputConfiguration();
// state queries
typedef int32_t (InputDevice::*GetStateFunc)(uint32_t sourceMask, int32_t code);
int32_t getState(int32_t deviceId, uint32_t sourceMask, int32_t code,
GetStateFunc getStateFunc);
bool markSupportedKeyCodes(int32_t deviceId, uint32_t sourceMask, size_t numCodes,
const int32_t* keyCodes, uint8_t* outFlags);
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
};
/* Reads raw events from the event hub and processes them, endlessly. */
class InputReaderThread : public Thread {
public:
InputReaderThread(const sp<InputReaderInterface>& reader);
virtual ~InputReaderThread();
private:
sp<InputReaderInterface> mReader;
virtual bool threadLoop();
};
/* Represents the state of a single input device. */
class InputDevice {
public:
InputDevice(InputReaderContext* context, int32_t id, const String8& name);
~InputDevice();
inline InputReaderContext* getContext() { return mContext; }
inline int32_t getId() { return mId; }
inline const String8& getName() { return mName; }
inline uint32_t getSources() { return mSources; }
inline bool isIgnored() { return mMappers.isEmpty(); }
void dump(String8& dump);
void addMapper(InputMapper* mapper);
void configure();
void reset();
void process(const RawEvent* rawEvent);
void getDeviceInfo(InputDeviceInfo* outDeviceInfo);
int32_t getKeyCodeState(uint32_t sourceMask, int32_t keyCode);
int32_t getScanCodeState(uint32_t sourceMask, int32_t scanCode);
int32_t getSwitchState(uint32_t sourceMask, int32_t switchCode);
bool markSupportedKeyCodes(uint32_t sourceMask, size_t numCodes,
const int32_t* keyCodes, uint8_t* outFlags);
int32_t getMetaState();
inline const InputDeviceCalibration& getCalibration() {
return mCalibration;
}
private:
InputReaderContext* mContext;
int32_t mId;
Vector<InputMapper*> mMappers;
String8 mName;
uint32_t mSources;
typedef int32_t (InputMapper::*GetStateFunc)(uint32_t sourceMask, int32_t code);
int32_t getState(uint32_t sourceMask, int32_t code, GetStateFunc getStateFunc);
InputDeviceCalibration mCalibration;
};
/* An input mapper transforms raw input events into cooked event data.
* A single input device can have multiple associated input mappers in order to interpret
* different classes of events.
*/
class InputMapper {
public:
InputMapper(InputDevice* device);
virtual ~InputMapper();
inline InputDevice* getDevice() { return mDevice; }
inline int32_t getDeviceId() { return mDevice->getId(); }
inline const String8 getDeviceName() { return mDevice->getName(); }
inline InputReaderContext* getContext() { return mContext; }
inline InputReaderPolicyInterface* getPolicy() { return mContext->getPolicy(); }
inline InputDispatcherInterface* getDispatcher() { return mContext->getDispatcher(); }
inline EventHubInterface* getEventHub() { return mContext->getEventHub(); }
virtual uint32_t getSources() = 0;
virtual void populateDeviceInfo(InputDeviceInfo* deviceInfo);
virtual void dump(String8& dump);
virtual void configure();
virtual void reset();
virtual void process(const RawEvent* rawEvent) = 0;
virtual int32_t getKeyCodeState(uint32_t sourceMask, int32_t keyCode);
virtual int32_t getScanCodeState(uint32_t sourceMask, int32_t scanCode);
virtual int32_t getSwitchState(uint32_t sourceMask, int32_t switchCode);
virtual bool markSupportedKeyCodes(uint32_t sourceMask, size_t numCodes,
const int32_t* keyCodes, uint8_t* outFlags);
virtual int32_t getMetaState();
protected:
InputDevice* mDevice;
InputReaderContext* mContext;
bool applyStandardPolicyActions(nsecs_t when, int32_t policyActions);
};
class SwitchInputMapper : public InputMapper {
public:
SwitchInputMapper(InputDevice* device);
virtual ~SwitchInputMapper();
virtual uint32_t getSources();
virtual void process(const RawEvent* rawEvent);
virtual int32_t getSwitchState(uint32_t sourceMask, int32_t switchCode);
private:
void processSwitch(nsecs_t when, int32_t switchCode, int32_t switchValue);
};
class KeyboardInputMapper : public InputMapper {
public:
KeyboardInputMapper(InputDevice* device, int32_t associatedDisplayId, uint32_t sources,
int32_t keyboardType);
virtual ~KeyboardInputMapper();
virtual uint32_t getSources();
virtual void populateDeviceInfo(InputDeviceInfo* deviceInfo);
virtual void dump(String8& dump);
virtual void reset();
virtual void process(const RawEvent* rawEvent);
virtual int32_t getKeyCodeState(uint32_t sourceMask, int32_t keyCode);
virtual int32_t getScanCodeState(uint32_t sourceMask, int32_t scanCode);
virtual bool markSupportedKeyCodes(uint32_t sourceMask, size_t numCodes,
const int32_t* keyCodes, uint8_t* outFlags);
virtual int32_t getMetaState();
private:
Mutex mLock;
struct KeyDown {
int32_t keyCode;
int32_t scanCode;
};
int32_t mAssociatedDisplayId;
uint32_t mSources;
int32_t mKeyboardType;
struct LockedState {
Vector<KeyDown> keyDowns; // keys that are down
int32_t metaState;
nsecs_t downTime; // time of most recent key down
} mLocked;
void initializeLocked();
bool isKeyboardOrGamepadKey(int32_t scanCode);
void processKey(nsecs_t when, bool down, int32_t keyCode, int32_t scanCode,
uint32_t policyFlags);
void applyPolicyAndDispatch(nsecs_t when, uint32_t policyFlags,
bool down, int32_t keyCode, int32_t scanCode, int32_t metaState, nsecs_t downTime);
ssize_t findKeyDownLocked(int32_t scanCode);
};
class TrackballInputMapper : public InputMapper {
public:
TrackballInputMapper(InputDevice* device, int32_t associatedDisplayId);
virtual ~TrackballInputMapper();
virtual uint32_t getSources();
virtual void populateDeviceInfo(InputDeviceInfo* deviceInfo);
virtual void dump(String8& dump);
virtual void reset();
virtual void process(const RawEvent* rawEvent);
virtual int32_t getScanCodeState(uint32_t sourceMask, int32_t scanCode);
private:
// Amount that trackball needs to move in order to generate a key event.
static const int32_t TRACKBALL_MOVEMENT_THRESHOLD = 6;
Mutex mLock;
int32_t mAssociatedDisplayId;
struct Accumulator {
enum {
FIELD_BTN_MOUSE = 1,
FIELD_REL_X = 2,
FIELD_REL_Y = 4
};
uint32_t fields;
bool btnMouse;
int32_t relX;
int32_t relY;
inline void clear() {
fields = 0;
}
} mAccumulator;
float mXScale;
float mYScale;
float mXPrecision;
float mYPrecision;
struct LockedState {
bool down;
nsecs_t downTime;
} mLocked;
void initializeLocked();
void sync(nsecs_t when);
void applyPolicyAndDispatch(nsecs_t when, int32_t motionEventAction,
PointerCoords* pointerCoords, nsecs_t downTime);
};
class TouchInputMapper : public InputMapper {
public:
TouchInputMapper(InputDevice* device, int32_t associatedDisplayId);
virtual ~TouchInputMapper();
virtual uint32_t getSources();
virtual void populateDeviceInfo(InputDeviceInfo* deviceInfo);
virtual void dump(String8& dump);
virtual void configure();
virtual void reset();
virtual int32_t getKeyCodeState(uint32_t sourceMask, int32_t keyCode);
virtual int32_t getScanCodeState(uint32_t sourceMask, int32_t scanCode);
virtual bool markSupportedKeyCodes(uint32_t sourceMask, size_t numCodes,
const int32_t* keyCodes, uint8_t* outFlags);
protected:
Mutex mLock;
struct VirtualKey {
int32_t keyCode;
int32_t scanCode;
uint32_t flags;
// computed hit box, specified in touch screen coords based on known display size
int32_t hitLeft;
int32_t hitTop;
int32_t hitRight;
int32_t hitBottom;
inline bool isHit(int32_t x, int32_t y) const {
return x >= hitLeft && x <= hitRight && y >= hitTop && y <= hitBottom;
}
};
// Raw data for a single pointer.
struct PointerData {
uint32_t id;
int32_t x;
int32_t y;
int32_t pressure;
int32_t touchMajor;
int32_t touchMinor;
int32_t toolMajor;
int32_t toolMinor;
int32_t orientation;
};
// Raw data for a collection of pointers including a pointer id mapping table.
struct TouchData {
uint32_t pointerCount;
PointerData pointers[MAX_POINTERS];
BitSet32 idBits;
uint32_t idToIndex[MAX_POINTER_ID + 1];
void copyFrom(const TouchData& other) {
pointerCount = other.pointerCount;
idBits = other.idBits;
for (uint32_t i = 0; i < pointerCount; i++) {
pointers[i] = other.pointers[i];
int id = pointers[i].id;
idToIndex[id] = other.idToIndex[id];
}
}
inline void clear() {
pointerCount = 0;
idBits.clear();
}
};
int32_t mAssociatedDisplayId;
// Immutable configuration parameters.
struct Parameters {
bool useBadTouchFilter;
bool useJumpyTouchFilter;
bool useAveragingTouchFilter;
} mParameters;
// Immutable calibration parameters in parsed form.
struct Calibration {
// Touch Area
enum TouchAreaCalibration {
TOUCH_AREA_CALIBRATION_DEFAULT,
TOUCH_AREA_CALIBRATION_NONE,
TOUCH_AREA_CALIBRATION_GEOMETRIC,
TOUCH_AREA_CALIBRATION_PRESSURE,
};
TouchAreaCalibration touchAreaCalibration;
// Tool Area
enum ToolAreaCalibration {
TOOL_AREA_CALIBRATION_DEFAULT,
TOOL_AREA_CALIBRATION_NONE,
TOOL_AREA_CALIBRATION_GEOMETRIC,
TOOL_AREA_CALIBRATION_LINEAR,
};
ToolAreaCalibration toolAreaCalibration;
bool haveToolAreaLinearScale;
float toolAreaLinearScale;
bool haveToolAreaLinearBias;
float toolAreaLinearBias;
bool haveToolAreaIsSummed;
int32_t toolAreaIsSummed;
// Pressure
enum PressureCalibration {
PRESSURE_CALIBRATION_DEFAULT,
PRESSURE_CALIBRATION_NONE,
PRESSURE_CALIBRATION_PHYSICAL,
PRESSURE_CALIBRATION_AMPLITUDE,
};
enum PressureSource {
PRESSURE_SOURCE_DEFAULT,
PRESSURE_SOURCE_PRESSURE,
PRESSURE_SOURCE_TOUCH,
};
PressureCalibration pressureCalibration;
PressureSource pressureSource;
bool havePressureScale;
float pressureScale;
// Size
enum SizeCalibration {
SIZE_CALIBRATION_DEFAULT,
SIZE_CALIBRATION_NONE,
SIZE_CALIBRATION_NORMALIZED,
};
SizeCalibration sizeCalibration;
// Orientation
enum OrientationCalibration {
ORIENTATION_CALIBRATION_DEFAULT,
ORIENTATION_CALIBRATION_NONE,
ORIENTATION_CALIBRATION_INTERPOLATED,
};
OrientationCalibration orientationCalibration;
} mCalibration;
// Raw axis information from the driver.
struct RawAxes {
RawAbsoluteAxisInfo x;
RawAbsoluteAxisInfo y;
RawAbsoluteAxisInfo pressure;
RawAbsoluteAxisInfo touchMajor;
RawAbsoluteAxisInfo touchMinor;
RawAbsoluteAxisInfo toolMajor;
RawAbsoluteAxisInfo toolMinor;
RawAbsoluteAxisInfo orientation;
} mRawAxes;
// Current and previous touch sample data.
TouchData mCurrentTouch;
TouchData mLastTouch;
// The time the primary pointer last went down.
nsecs_t mDownTime;
struct LockedState {
Vector<VirtualKey> virtualKeys;
// The surface orientation and width and height set by configureSurfaceLocked().
int32_t surfaceOrientation;
int32_t surfaceWidth, surfaceHeight;
// Translation and scaling factors, orientation-independent.
int32_t xOrigin;
float xScale;
float xPrecision;
int32_t yOrigin;
float yScale;
float yPrecision;
float geometricScale;
float toolAreaLinearScale;
float toolAreaLinearBias;
float pressureScale;
float sizeScale;
float orientationScale;
// Oriented motion ranges for input device info.
struct OrientedRanges {
InputDeviceInfo::MotionRange x;
InputDeviceInfo::MotionRange y;
bool havePressure;
InputDeviceInfo::MotionRange pressure;
bool haveSize;
InputDeviceInfo::MotionRange size;
bool haveTouchArea;
InputDeviceInfo::MotionRange touchMajor;
InputDeviceInfo::MotionRange touchMinor;
bool haveToolArea;
InputDeviceInfo::MotionRange toolMajor;
InputDeviceInfo::MotionRange toolMinor;
bool haveOrientation;
InputDeviceInfo::MotionRange orientation;
} orientedRanges;
// Oriented dimensions and precision.
float orientedSurfaceWidth, orientedSurfaceHeight;
float orientedXPrecision, orientedYPrecision;
struct CurrentVirtualKeyState {
bool down;
nsecs_t downTime;
int32_t keyCode;
int32_t scanCode;
} currentVirtualKey;
} mLocked;
virtual void configureParameters();
virtual void dumpParameters(String8& dump);
virtual void configureRawAxes();
virtual void dumpRawAxes(String8& dump);
virtual bool configureSurfaceLocked();
virtual void dumpSurfaceLocked(String8& dump);
virtual void configureVirtualKeysLocked();
virtual void dumpVirtualKeysLocked(String8& dump);
virtual void parseCalibration();
virtual void resolveCalibration();
virtual void dumpCalibration(String8& dump);
enum TouchResult {
// Dispatch the touch normally.
DISPATCH_TOUCH,
// Do not dispatch the touch, but keep tracking the current stroke.
SKIP_TOUCH,
// Do not dispatch the touch, and drop all information associated with the current stoke
// so the next movement will appear as a new down.
DROP_STROKE
};
void syncTouch(nsecs_t when, bool havePointerIds);
private:
/* Maximum number of historical samples to average. */
static const uint32_t AVERAGING_HISTORY_SIZE = 5;
/* Slop distance for jumpy pointer detection.
* The vertical range of the screen divided by this is our epsilon value. */
static const uint32_t JUMPY_EPSILON_DIVISOR = 212;
/* Number of jumpy points to drop for touchscreens that need it. */
static const uint32_t JUMPY_TRANSITION_DROPS = 3;
static const uint32_t JUMPY_DROP_LIMIT = 3;
/* Maximum squared distance for averaging.
* If moving farther than this, turn of averaging to avoid lag in response. */
static const uint64_t AVERAGING_DISTANCE_LIMIT = 75 * 75;
struct AveragingTouchFilterState {
// Individual history tracks are stored by pointer id
uint32_t historyStart[MAX_POINTERS];
uint32_t historyEnd[MAX_POINTERS];
struct {
struct {
int32_t x;
int32_t y;
int32_t pressure;
} pointers[MAX_POINTERS];
} historyData[AVERAGING_HISTORY_SIZE];
} mAveragingTouchFilter;
struct JumpyTouchFilterState {
uint32_t jumpyPointsDropped;
} mJumpyTouchFilter;
struct PointerDistanceHeapElement {
uint32_t currentPointerIndex : 8;
uint32_t lastPointerIndex : 8;
uint64_t distance : 48; // squared distance
};
void initializeLocked();
TouchResult consumeOffScreenTouches(nsecs_t when, uint32_t policyFlags);
void dispatchTouches(nsecs_t when, uint32_t policyFlags);
void dispatchTouch(nsecs_t when, uint32_t policyFlags, TouchData* touch,
BitSet32 idBits, uint32_t changedId, uint32_t pointerCount,
int32_t motionEventAction);
void applyPolicyAndDispatchVirtualKey(nsecs_t when, uint32_t policyFlags,
int32_t keyEventAction, int32_t keyEventFlags,
int32_t keyCode, int32_t scanCode, nsecs_t downTime);
bool isPointInsideSurfaceLocked(int32_t x, int32_t y);
const VirtualKey* findVirtualKeyHitLocked(int32_t x, int32_t y);
bool applyBadTouchFilter();
bool applyJumpyTouchFilter();
void applyAveragingTouchFilter();
void calculatePointerIds();
};
class SingleTouchInputMapper : public TouchInputMapper {
public:
SingleTouchInputMapper(InputDevice* device, int32_t associatedDisplayId);
virtual ~SingleTouchInputMapper();
virtual void reset();
virtual void process(const RawEvent* rawEvent);
protected:
virtual void configureRawAxes();
private:
struct Accumulator {
enum {
FIELD_BTN_TOUCH = 1,
FIELD_ABS_X = 2,
FIELD_ABS_Y = 4,
FIELD_ABS_PRESSURE = 8,
FIELD_ABS_TOOL_WIDTH = 16
};
uint32_t fields;
bool btnTouch;
int32_t absX;
int32_t absY;
int32_t absPressure;
int32_t absToolWidth;
inline void clear() {
fields = 0;
}
} mAccumulator;
bool mDown;
int32_t mX;
int32_t mY;
int32_t mPressure;
int32_t mToolWidth;
void initialize();
void sync(nsecs_t when);
};
class MultiTouchInputMapper : public TouchInputMapper {
public:
MultiTouchInputMapper(InputDevice* device, int32_t associatedDisplayId);
virtual ~MultiTouchInputMapper();
virtual void reset();
virtual void process(const RawEvent* rawEvent);
protected:
virtual void configureRawAxes();
private:
struct Accumulator {
enum {
FIELD_ABS_MT_POSITION_X = 1,
FIELD_ABS_MT_POSITION_Y = 2,
FIELD_ABS_MT_TOUCH_MAJOR = 4,
FIELD_ABS_MT_TOUCH_MINOR = 8,
FIELD_ABS_MT_WIDTH_MAJOR = 16,
FIELD_ABS_MT_WIDTH_MINOR = 32,
FIELD_ABS_MT_ORIENTATION = 64,
FIELD_ABS_MT_TRACKING_ID = 128,
FIELD_ABS_MT_PRESSURE = 256,
};
uint32_t pointerCount;
struct Pointer {
uint32_t fields;
int32_t absMTPositionX;
int32_t absMTPositionY;
int32_t absMTTouchMajor;
int32_t absMTTouchMinor;
int32_t absMTWidthMajor;
int32_t absMTWidthMinor;
int32_t absMTOrientation;
int32_t absMTTrackingId;
int32_t absMTPressure;
inline void clear() {
fields = 0;
}
} pointers[MAX_POINTERS + 1]; // + 1 to remove the need for extra range checks
inline void clear() {
pointerCount = 0;
pointers[0].clear();
}
} mAccumulator;
void initialize();
void sync(nsecs_t when);
};
Native input dispatch rewrite work in progress. The old dispatch mechanism has been left in place and continues to be used by default for now. To enable native input dispatch, edit the ENABLE_NATIVE_DISPATCH constant in WindowManagerPolicy. Includes part of the new input event NDK API. Some details TBD. To wire up input dispatch, as the ViewRoot adds a window to the window session it receives an InputChannel object as an output argument. The InputChannel encapsulates the file descriptors for a shared memory region and two pipe end-points. The ViewRoot then provides the InputChannel to the InputQueue. Behind the scenes, InputQueue simply attaches handlers to the native PollLoop object that underlies the MessageQueue. This way MessageQueue doesn't need to know anything about input dispatch per-se, it just exposes (in native code) a PollLoop that other components can use to monitor file descriptor state changes. There can be zero or more targets for any given input event. Each input target is specified by its input channel and some parameters including flags, an X/Y coordinate offset, and the dispatch timeout. An input target can request either synchronous dispatch (for foreground apps) or asynchronous dispatch (fire-and-forget for wallpapers and "outside" targets). Currently, finding the appropriate input targets for an event requires a call back into the WindowManagerServer from native code. In the future this will be refactored to avoid most of these callbacks except as required to handle pending focus transitions. End-to-end event dispatch mostly works! To do: event injection, rate limiting, ANRs, testing, optimization, etc. Change-Id: I8c36b2b9e0a2d27392040ecda0f51b636456de25
2010-04-23 01:58:52 +00:00
} // namespace android
#endif // _UI_INPUT_READER_H